> Hi all,
> My typical backup opernations will involve backing up a fileserver
> (12GB), web/mail server (2GB), and database server (Informix SE, 5 GB).
> All three servers are on redhat 5.2 and I want to back them all through
> NFS at one place. Occassionally, I may have to backup some windows
> machines on the network. I've samba running on the fileserver. I've an
> Exabyte mammoth tape drive with AdvanSys Ultra wide SCSI card.
I have a similar situation where I use a linux machine
with a 14 tape Quantum DLT tape robot to do daily incremental
backups of both unix machines via NFS, and and NT machines
via samba.
In my case, two of the disk servers are 70GB raids, both
hovering at 90% full. One is an NT, one linux.
And a few other unix machines, a mixture of SGIs and linux.
It sounds like you don't mind 'rolling your own' backup script,
so you might want to investigate my technique.
I use gnutar to backup the NFS volumes, and samba smbclient(1)'s
tar abilities to get the NT server. This way all the data is
backed up using tar archives.
The gnutar command line is basically:
nice tar -c \
-v \
-f /dev/nst0 \
--same-order \
--new-volume-script ./next-tape \
--label ${tapename}_${magazine_name} \
--multi-volume \
--one-file-system \
--listed-incremental ./snapshot_${tapename}_${magazine_name} \
$dirs >& $error_log
..and the smbclient command is (basically):
nice smbclient //ntserver/files \
$backup_passwd \
-U $backup_user \
-W $our_workgroup \
-TbcS 20 /dev/nst0 ./next-tape $dirs >& $error_log
Note: You won't see the smbclient 'S' flag in the man page;
I hacked that into smbclient so that I could tell it to run
my './next-tape' script to load the next tape when the end of
a tape is encountered. I need to publish the diff(1)s I made
so that other people can use that. I love open source ;)
But it doesn't sound like you have a tape robot to worry about,
so that's probably a non issue for you.
Also note: I found that gnutar has a subtle bug with
incrementals, where after a full backup is made, files that
are *renamed* are not caught in the incrementals. In our case,
this was a negligable enough problem, so we ignore it, but
anxiously await a fix from the gnu maintainers (I reported it
in fall 98).
My backups script has alot of *around the above commands
to make sure the directories are mounted, the tape is properly
positioned relative to the last time the script was run, keeping
track of which magazine/tape number, etc. Simple admin-y stuff.
If you want to roll your own, I can say the above has been working
for me the last 3 or 4 months now. And yes, we've had to do a few
restores, including a complete raid recovery 8^O which actually
worked, thank god.
Quote:> [..] BRU [..]
No opinion on BRU, sorry.
Hope the recommendations above help.
--
/\_/\
|o,o|
\/ )
----mm---------------------------------------------------------------------
Greg Ercolano UNIX NightOwl / Systems Programmer & Admin
---------------------------------------------------------------------------
by